The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute complex shell commands using msfconsole and msfvenom. These commands are designed to exploit vulnerabilities in operating systems (e.g., MS17-010), databases (PostgreSQL, MySQL, MSSQL), and web services.
[PRIVILEGE_ESCALATION]: Instructions are provided for gaining elevated privileges on target systems using Meterpreter commands like getsystem, hashdump, and creds_all (via the Kiwi extension), as well as lateral movement techniques like PsExec.
[PERSISTENCE]: The skill describes methods to maintain long-term access to compromised targets using multiple transports (TCP/HTTP) and configuring handlers with ExitOnSession false to ensure sessions remain active across reconnections.
[DATA_EXFILTRATION]: Instructions include commands to exfiltrate data from target systems to the agent's local environment, specifically using Meterpreter's download command (e.g., download C:\flag.txt /tmp/flag.txt).
[DYNAMIC_EXECUTION]: The skill utilizes msfvenom to dynamically generate executable payloads (EXE, ELF, WAR, Python, PHP) at runtime, which are then intended for deployment and execution on target machines.
[INDIRECT_PROMPT_INJECTION]: The skill possesses a vulnerability surface for indirect prompt injection by reading untrusted output from interactive sessions.
Ingestion points: Untrusted data enters the context through interactive_session(action="read", ...) which captures the output of commands executed on remote targets.
Boundary markers: The instructions lack explicit delimiters or instructions to ignore potential commands embedded in target output.
Capability inventory: The skill has access to bash for executing MSF tools and the interactive_session MCP tool for sending commands to active shells.
Sanitization: There is no evidence of output sanitization or validation before the agent processes the data to determine its next action.

msf-oneshot